Can you think of any other disadvantages?
You could call it either way to be honest, but if you need pixel-perfect collision (ie based on image/texture overlap rather than primitive collision) then you'll have to roll your own system whereas in 2D you'd simply employ the existing function.